Direct Bandgap Germanium for Active Silicon Photonics Application
Machine learning and data centers require data interconnects with orders of magnitude more bandwidth. This could be achieved using optical interconnects, but the Group IV semiconductors compatible with silicon fabrication (Si, Ge, C, and Sn) emit almost no light. All Group IV elements have an indirect bandgap that prevents light emission but applying strong tensile strain to Ge or alloying it with C or Sn could induce a direct bandgap. This dissertation focused on three such projects. In the first project, I successfully incorporated C in Ge using CBr4 as a C precursor. Reflection high energy electron diffraction (RHEED) and atomic force microscopy (AFM) showed better surface reconstruction and lower surface roughness for the lowest growth temperature, Tsub = 215 °C. Increasing growth temperatures turned the RHEED pattern spotty and increased the surface roughness. High resolution x-ray diffraction (HRXRD) confirmed the RHEED and AFM results, showing better crystal quality at Tsub = 215 °C, with a higher-angle peak corresponding to tensile strain from a nominally substitutional carbon content of 0.71%. We believe this is a lower limit for the total composition of C in the sample, because ab-initio simulations showed that Ge vacancies could not explain the peak shift, and C interstitials would push the lattice toward compressive strain. Raman spectroscopy showed a clear Ge-C local mode at 530 cm-1 for growths from 215-324 °C, confirming the substitutional carbon incorporation in germanium. In contrast with previous reports of Ge:C growth, these samples showed no amorphous or graphitic carbon in Raman. Furthermore, these samples produced the first reported photoluminescence (PL) below the Ge bandgap, near 0.61 eV at 83 K, in agreement with band anti-crossing and computational models. In the second project, I added a beam of atomic H to the Ge1-xCx growth to reduce undesirable C-C bonds and similar C clusters on the growth surface. Unlike the H-free growths, these samples showed the smoothest surface and narrowest XRD linewidth at higher temperatures: Tsub = 324 °C. More significantly, Raman spectroscopy showed a 4x stronger Ge-C local mode peak intensity compared to the samples grown without H, which suggests a much larger fraction of C substitutional in the lattice. Again, PL showed emission below the Ge bandgap in these samples, near 0.60 eV at 83 K. Finally, in the third project, I modeled ridge waveguide lasers using tensile strained Ge for the active region, with stress provided by SiNx stress liners. I performed 2.5D mechanical stress modeling in COMSOL Multiphysics, then combined the strain profile with ab-initio data to produce gain/absorption and refractive index profiles across the waveguide. From these, I calculated the optical mode profile, modal gain, and material losses in the laser. Intervalence band absorption loss was found to be the dominant loss. The threshold current density was found to be 1.4 kA/cm2, which was almost 10× higher than typical GaAs-based lasers. Therefore, strain alone is insufficient to produce efficient lasers, although it could aid in producing a direct bandgap from Ge alloys. Together, these results offer a route to lasers and other active photonic devices on silicon.
Group IV, Germanium, Carbon, Germanium carbide, Tensile germanium, Germanium waveguide, MBE, CBr4
Reza, S. (2022). Direct bandgap germanium for active silicon photonics application (Unpublished dissertation). Texas State University, San Marcos, Texas.